Histograms, with night photography examples

[DanK]

In a recent thread, there was a lot of discussion about interpreting histograms, in general and in night photography specifically. I thought some concrete examples might help people who are still puzzling over this.

First, if you are really puzzled by histograms, read the tutorial on this site first. That will make what follows far easier to understand.

Second, almost nothing I write here is specific to night photography. Where something is specific to night photography, I'll mention it.

First, I'll start with this shot of the Boston Christian Science center, along with its histogram:





The first thing to note is that the histogram is unimodal (one hump). This is because most of the pixels are in that narrow range of luminance, and there are only a few at more extreme values. This is a result of the lighting, not my exposure. Second, the mode (the peak) is toward the dark end. This is a deliberate choice, which I could make either in exposure or postprocessing: I wanted the image a little on the dark side, because it is a night photograph. In fact, I exposed brighter than this, exposing as much as I could without having anything but the lights clipped at the top, and then darkened in post because that creates less noise. Third, it's righ-skewed--a longer tail on the right than the left. This is because there are some pixels, not many, spread out all the way from the hump to the brightest value.

Now let's look at an original capture of a pond not far from my house at night, along with its histogram:





This one is bimodal (two humps). The lower one is all the pixels in the very dark areas, while the top one is the pixels in the much brighter sky. There isn't much in between, in terms of luminance, which is why there is the long, low expanse between the two humps.

You can see that there is one very undesirable characteristic of this histogram: a lot of the image is in a very dark part of the histogram, which means that the ratio of signal (which varies with brightness) to noise (which doesn't) will be low. I did this because I didn't want to clip at the high end, and you can see that I was right on the verge of doing so.

Here's the final image, with its histogram:





You can see that I brought the two modes closer together: I moved the shadows up and the highlights down. Once again, the mass of the final histogram is to the left, but that is because it is a night photo and ought to look dark.

Finally, here's one with a multimodal histogram. This is the final image, but in this case, the original wasn't dramatically different.





Once again, the mass is toward the left--it's right-skewed--because it is night photograph, and I wanted the average luminance to be low. I could have shifted everything rightward, at the cost of some clipping, if I wanted it to look more like daytime. The left-most mode is the area of trees that is nearly black.

I hope this is instructive. If people would find it useful, it wouldn't be hard to show a similar set with daytime photos.

[Manfred M]

Great post Dan.

If you happen to have one, it might be interesting to show one of the images and the histogram with PP work done to it, just to show how that step does have an impact on the histogram.

[DanK]

If you happen to have one, it might be interesting to show one of the images and the histogram with PP work done to it, just to show how that step does have an impact on the histogram.

There's one pair in the original post, but here is another that might be more instructive. Here's the original capture of the Christian Science Center image at the top of the first post, along with its histogram:





The original capture is considerably brighter than in the final product, and the peak of the histogram is accordingly farther to the right. That's as bright as I could make the exposure without clipping details I wanted to preserve. I then darkened the image to look more light it did at night. The ugly yellow color is characteristic of sodium vapor lamps.

The size of the mass around the mode indicates that a large proportion of the image has fairly similar luminance.

That this shows is that there is not anything inherent in night photos that creates a big mass at the left of the histogram. It depends on the mix of luminance in the image and how long you expose.